The shape of the tibial spines as imaging biomarker for incident knee osteoarthritis

Abstract

Objective

Morphological aspects of the tibial spines on radiography have received little attention in the evaluation of disease status or disease prediction in OA. This study explored the inter-observer reliability and associations with clinical and radiographic knee OA incidence for medial and lateral spiking, the medial and lateral edge angles, and the height of the tibial spines.

Design

Data from 344 middle-aged women who were overweight/obese and free of radiographic OA and knee symptoms, was used for the analyses. Outcomes were the incidence of radiographic knee OA and clinical knee OA after 30 months. Generalized Estimating Equations, to adjust for multiple measures within individuals, and a backward selection method (p > 0.1 for removal) were used to obtain prediction models. For each outcome, AUC values for the remaining factors and 95 % CI were calculated.

Results

Apart from the lateral spine edge angle, all features had moderate to excellent reliability. Only the lateral spine angle showed a significant association with incident radiographic knee OA (AUC of 0.63 (95 % CI 0.53–0.73)). Medial spiking, lateral spiking, and the medial edge angle of the medial tibial spine were significantly associated with incident clinical knee OA (AUC of 0.71 (95 % CI 0.62–0.81).

Conclusions

Morphological aspects of the tibial spines can reliably be obtained from radiography and should be considered in future prediction models, combined with other known predictors from patient history, physical examination, and/or imaging.